Visualizing Time-varying Twitter Data with SentimentClock - LIRMM - Laboratoire d’Informatique, de Robotique et de Microélectronique de Montpellier Access content directly
Conference Poster Year : 2014

Visualizing Time-varying Twitter Data with SentimentClock

Abstract

Temporal dimension contains important information for sentiment analysis of microblog data such as tweets. Previous works on sentiment visualization could not address the multidimensional nature of sentiment together with temporal information. In this work, we introduce SentimentClock for visualizing the sentiment of time-varying Twitter data on 2D affective space. Our visualization enables various interesting tasks : (1) Visualize and compare temporal variations of sentiments. (2) Compare sentiments variations of tweets on different topics. (3) Visualize the distribution of tweets on 2D affective space. (4) Visualize both dimensions of sentiments (i.e. valence, arousal) and their semantic meanings (e.g. elated, stressed). Fig.1 SentimentClock of the tweets collected on 2013 Australian election day (7-Sep-2013) Fig.2 SentimentClocks of tweets on two different topics: Australian Politics (left) and World Cup 2014 (right) Fig.1 shows the sentiment visualization of 36016 related tweets posted on 2013 Australian election day. In the evening (18:00 to 22:00), which is the vote counting and result releasing period, tweets are found to have both high arousal and valence, primarily falling into the elated and excited range with high strength. Fig.2 shows the sentiment visualization of 71200 tweets on two topics. Tweets on the topic " Australian Politics " are more spread out along the sentiment wheel and express more negative sentiments, e.g. upset and stressed. However, tweets on the topic " World cup 2014 " are mainly concentrated within the range of content and elated.
Fichier principal
Vignette du fichier
sentimentclockb.pdf (1.2 Mo) Télécharger le fichier
sentimentclocka.pdf (891.14 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)
Loading...

Dates and versions

lirmm-01275390 , version 1 (17-02-2016)

Identifiers

  • HAL Id : lirmm-01275390 , version 1

Cite

Florence Ying Wang, Arnaud Sallaberry, Karsten Klein, Masahiro Takatsuka. Visualizing Time-varying Twitter Data with SentimentClock. InfoVis: Information Visualization, 2014, Paris, France. IEEE Information Visualization Conference (InfoVis 2014), Poster Abstracts of IEEE VisWeek 2014, 2014. ⟨lirmm-01275390⟩
134 View
348 Download

Share

Gmail Facebook X LinkedIn More